CN106964137B - A kind of ball service behavior rating method based on image procossing - Google Patents
A kind of ball service behavior rating method based on image procossing Download PDFInfo
- Publication number
- CN106964137B CN106964137B CN201710100439.5A CN201710100439A CN106964137B CN 106964137 B CN106964137 B CN 106964137B CN 201710100439 A CN201710100439 A CN 201710100439A CN 106964137 B CN106964137 B CN 106964137B
- Authority
- CN
- China
- Prior art keywords
- service
- image
- ball
- coordinate
- mark point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 230000033001 locomotion Effects 0.000 claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 13
- 239000003086 colorant Substances 0.000 claims description 7
- 230000000717 retained effect Effects 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 4
- 210000000707 wrist Anatomy 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 239000003550 marker Substances 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 claims 1
- 238000012549 training Methods 0.000 abstract description 17
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000000386 athletic effect Effects 0.000 abstract description 2
- 230000000153 supplemental effect Effects 0.000 abstract description 2
- 238000012545 processing Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000008054 signal transmission Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B69/00—Training appliances or apparatus for special sports
- A63B69/38—Training appliances or apparatus for special sports for tennis
- A63B69/385—Training appliances or apparatus for special sports for tennis for practising the serve
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physical Education & Sports Medicine (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The ball service behavior rating method based on image procossing that the invention discloses a kind of, comprising the following steps: step 1: color mark is carried out to the artis of service arm, acquires ball service image；Step 2: by ball service image configuration dictionary, reconstructing glitch-free service figure；Step 3: for the service figure of acquisition, carrying out sport foreground extraction；Step 4: extracting mark point for sport foreground, and dot profile lookup is marked, surround profile with smallest circle, return to central coordinate of circle as body joint point coordinate；Step 5: being trained by the body joint point coordinate data that step 1~4 are extracted, realize to the classification of service track, establish assessment system；Step 6: service coordinate to be measured being acquired by the method for step 1~4, the assessment system that input step (5) obtains is evaluated；The method of the present invention realizes technology action analysis when tennis player's service, improves athletic training efficiency, Motion Technology is improved, to achieve the purpose that supplemental training.
Description
Technical field
The invention belongs to field of image processing more particularly to a kind of ball service behavior rating sides based on image procossing
Method.
Background technique
Human motion analysis based on video is an important research direction of computer vision field, it is from video sequence
Detect moving object in column, extract human body key position, obtain the useful information of human motion, realize to human action,
The further analysis and identification of posture etc..
Traditional tennis service training mode is inveteracy to be present among present training.Sportsman is come
It says, must just carry out prolonged, white silk repeatedly according to the instruction of coach to the essential of exercise that expertly masters a skill
It practises.The training method based on experience is used for a long time, and sportsman's technical movements are instructed and supervised only according to by means of coach
It superintends and directs, there is subjectivity, such case seriously constrains the raising of tennis level.
In order to improve result of training, also there is researcher to propose some service training devices, such as Publication No.
The patent document of CN203469419U discloses shuttlecock service training system, is related to a kind of shuttlecock training system.Be in order to
Adapt to the demand that badminton service robot moves the intelligent training of property, accuracy.Array indicator light is distributed in the one of shuttlecock court
Side, the control signal input of array indicator light and the control signal output of control circuit connect；IR signal reception electricity
The infrared signal output end on road and the infrared signal input terminal of control circuit connect；Infrared signal receiving circuit is infrared for receiving
The infrared signal of signal transmission circuit transmitting；The push button signalling output end of array key and the key of infrared signal transmission circuit
Signal input part connection；Video camera is used to acquire the image of service placement；The image signal output end and control circuit of video camera
Picture signal input terminal connection；The control signal output of control circuit and the control signal input of video camera connect.It should
Invention is suitable for shuttlecock service training process.
But above-mentioned serving trainer structure exists the problems such as structure is complicated, inconvenient for use and test inaccuracy, be because
This proposes a kind of tennis service behavior rating method based on image procossing for this problem, realize tennis training it is objective,
It effectively instructs very necessary.
Summary of the invention
The ball service behavior rating system based on image procossing that the present invention provides a kind of, by image processing techniques come
The flat service technology in ball is studied, ball training efficiency can be improved, assists ball training.
A kind of ball service behavior rating method based on image procossing, comprising the following steps:
Step 1: color mark being carried out to the artis of service arm, acquires ball service image；
Step 2: the ball service image configuration dictionary acquired by step 1 reconstructs glitch-free service figure；
Step 3: the service figure obtained for step 2 carries out sport foreground extraction；
Step 4: extracting mark point for the sport foreground after step 3 operation, and dot profile lookup is marked, with most
Roundlet surrounds profile, returns to central coordinate of circle as body joint point coordinate；
Step 5: it is trained by the body joint point coordinate data that step 1~4 are extracted, realizes the classification to service track,
Assessing ball service is successfully to serve a ball or unsuccessfully serve a ball, and establishes assessment system；
Step 6: service coordinate to be measured, the assessment system that input step (5) obtains are acquired by the method for step 1~4
It is evaluated.
It is ball to can be tennis, shuttlecock or vollyball etc..
Preferably, in step 1, color mark is carried out to the artis of service arm.To the service arm of each sportsman
Upper joint click through pedestrian be label, can when extracting target point more hard objectives point color characteristic, so as to foundation target point
Color characteristic preferably removes the interference of different colours, and after seeking advice from professional athlete, determines the presence of mark point to ball
Class service technique movement not substantive influence, acquiring ball service image, specific step is as follows:
1-1 selects the red marker color as artis, and in the point of label three A, B and the C of swinging the bat on arm, A represents wrist
Mark point close to racket part, B represent the mark point in elbow portion, and C represents service arm close to the mark point of shoulder；
High-speed camera is placed on the right side of experimental subjects by 1-2, apart from penalty mark 4m~6m, highly to adopt at 1.0m~1.6m
Collect ball service image.
Preferably, in step 2, dictionary is constructed by a series of noise-containing ball service figures that step 1 acquires, according to dilute
Thin expression reconstructs glitch-free service figure.Reservation while method denoising according to rarefaction representation reconstruct can effectively realize denoising
The marginal information of target to be detected, although the reconstruction of dictionary is than relatively time-consuming in rarefaction representation restructuring procedure, according to rarefaction representation
The dictionary that an image known to theory trains can characterize other similar images, it is contemplated that the rapidity of delivery of service, in order to make
The dictionary trained is more representative, and n (30 >=n >=10) the frame picture training service that may be selected in the primary service movement of interception is dynamic
The dictionary of work can be rebuild by rarefaction representation and be denoised dictionary application in entire video flowing later, instructed without repeating
It practises handwriting allusion quotation, solves the problems, such as that rarefaction representation reconstruct denoising real-time is poor.Specific step is as follows:
2-1 initialization: generate X=Y, initial dictionary D by DCT；
Above-mentioned X indicates the image after removal noise to be solved, its initial value is set as Y, Y is figure polluted by noise
Picture, the size of X, Y be N × N × 3 (3 indicate that the three primary colors of color images are red, green, the corresponding dimension of blue submatrix,
X, Y is color image), Y is divided into M image subblockSubimage block zjSize isAnd image subblock meets the sparse model z of triple (ε, s, D)j=D α, i.e., it is sub
Image block zjReconstruction error is ε, and code coefficient degree of rarefication is s, and dictionary is D ∈ R3n×K, x ∈ R3nFor signal z to be processedjAgain it arranges
The dimensional vector arranged, α are zjRarefaction representation on dictionary D.
Subimage block zjDenoising image block estimation model for example shown in formula (1):
μ in formulajFor the weight for controlling degree of rarefication and reconstruction error.
Formula (1) is modified slightly to obtain shown in the denoising Image estimation model such as formula (2) of entire image:
X is the clean image (image after denoising) recovered, sub-block x in formula (2)ij=RijX, Rij∈Rn×NBe
The matrix extracted in (i, j) a image block.Similarity between the clean figure of its recovery of first item constraint and noise pattern, Section 2
Prior-constrained for degree of rarefication, Section 3 is the constraint of sub-block reconstruction error.
2-2 fixes dictionary D, and changing formula (2) is formula (3), solves the code coefficient of each sub-block:
The factor alpha that 2-3 fixing step 2-2 is solvedij, formula (2) is turned to formula (4), dictionary D is solved:
Dictionary D is solved using K-SVD algorithm: first finding atom d in Dl(l=1 .., k) coefficient of correspondence αijUnder nonzero element
Mark set ωl=(i, j) | αij(l) ≠ 0 indexed set ω }, is then calculatedlIn each descend target residual error, such as formula (5):
By residual error setColumn vector constitute residual matrix El, SVD is carried out to it is decomposed into U Δ VT.That is U
One column update atom dl, V first row is multiplied with first singular value of Δ updates coefficientFinally obtain adaptive word
Allusion quotation D.
2-4 restores clean image X according to formula (6):
Preferably, in step 3, sport foreground extraction is carried out using mixed Gaussian background modeling.Mixed Gaussian background modeling
A kind of method of background description being built upon on the basis of pixel samples statistical information, it uses the system such as probability density of pixel
The meter sample value of information (such as: the expectation of each mode and standard deviation, the quantity of mode) within the quite a long time carrys out table
Show background.Then, object pixel is determined by statistics calculus of finite differences (such as 3 σ principles, σ are standard deviation), therefore according to mixed Gaussian
Background modeling extracts movement velocity of the method for prospect independent of object, adapts to the extraction of fast moving objects, is well suited for
The extraction of quick ball delivery of service.
Preferably, in step 4, mark point is extracted for the service figure after step 3 operation, and dot profile is marked and looks into
It looks for, surrounds profile with smallest circle, return to central coordinate of circle as body joint point coordinate.Mark point artificially selects, and color is it is known that benefit
The smallest circle that mark point is extracted with colouring information, can further eliminate picture noise, and is constituted with label dot profile
The center of circle can more accurately characterize mark point position as body joint point coordinate, because the center of circle is several profiles by being located on circle
What point determined together, exclusiveness is had more, smallest circle can guarantee that the profile point for carrying out center of circle extraction is more likely in mark point.
Specific step is as follows:
The rgb value of each of the foreground image that 4-1 traversal step 3 obtains pixel, for meeting mark point color
Being retained in range, obtains binary image f；
4-2 carries out profile to bianry image f and searches acquisition contour images f1；
4-3 is to image f1The wheel for separately including A, B, C tri- provided in step 1-1 labels is obtained by Hough transform
The smallest three circles of radius of exterior feature point, the center of circle of these three circles are required mark point.
Preferably, in step 4-1, the rgb value of each of foreground image that traversal step 3 obtains pixel, for
Meet being retained in mark point color gamut.According to known label point color information, traversing entire image can be by step 3
The noise left after processing further removes completely.Obtaining binary image f, specific step is as follows:
4-1-1 calculates the three primary colors pixel value of all pixels point in foreground image: I=2R-G-B, and will calculate
The value I of acquisition saves as new matrix a Z, i.e., new image Z；
The average value of all I is denoted as A in 4-1-2 calculating matrix Z；
All pixel values greater than A are set to 255 in 4-1-3 matrix I, and all pixel values less than or equal to A are set to 0, obtain
Binary image f.
Preferably, in step 5, the body joint point coordinate extracted to step 4 is trained, and realizes the classification to service track,
Assessing ball service is successfully to serve a ball or unsuccessfully serve a ball.Service track characterizes the motion profile of arm, can characterize service
Whether posture is correct, so can determine to be successfully to serve a ball or unsuccessfully serve a ball by service track.Specific steps are as follows:
5-1 distinguishes the X, Y coordinates of mark point of video A, B, C tri- labels of sequential storage service in each frame image, deposits
The coordinate of storage is the running track of mark point；
5-2 obtains N+M (N >=150, M >=50) by step 1 to step 4 and organizes motion profile, wherein service successful trail Ni
+MiGroup, service failure motion profile Nj+M jGroup, Ni+Nj=N, Ni> Nj, Mi+Mj=M, Mi> Mj；
5-3 is trained support vector machines using the N group data of step 5-2, then with the M group data of step 5-2 into
Row test, if service pass flag is 1, service fail flag is -1, completes the classification of service track.
By the application of rarefaction representation algorithm and mixed Gaussian background modeling method, can effectively overcome in ball service image
The influence of different colours noise effectively improves target detection accuracy.
Beneficial effects of the present invention:
The method of the present invention realizes technology action analysis when tennis player's service, helps coach and sportsman's discovery not
The movement or malfunction of specification improve athletic training efficiency, Motion Technology are improved, to achieve the purpose that supplemental training.
Detailed description of the invention
Fig. 1 is the process line frame graph of the method for the present invention.
Fig. 2 is the image of service arm mark point.
Fig. 3 is the video frame images for demonstrating treatment effect.
Fig. 4 is sparse denoising figure.
A~c of Fig. 5 is three foreground extraction figures that mixed Gaussian background modeling is moved.
The a and b of Fig. 6 is the binary map and profile diagram of the mark point obtained after handling motion foreground picture.
Specific embodiment
As shown in Figure 1, the method for building up of the tennis service behavior rating system based on image procossing of the present embodiment, including
Following steps:
Step 1: color mark is carried out to the artis of service arm, tennis service video is acquired by high-speed camera,
It is specific as follows:
Artis on sportsman's service arm is marked with red, as shown in Fig. 2, A point represents wrist close to ball
The mark point of part is clapped, B point represents the mark point in elbow portion, and C point represents service arm close to the mark point of shoulder.By IO
The Flare 2M360CCL high-speed camera of Industries company is placed on the right side of experimental subjects, high apart from penalty mark 4m~6m
Degree is acquisition service video at 1.0m~1.6m.
Step 2: constructing dictionary by a series of noise-containing tennis service figures acquired, reconstruct nothing according to rarefaction representation
The service figure of interference；The frame in the tennis service video of acquisition is chosen (for selected group of tennis service video sequence of acquisition
187th frame) for image as treatment process demonstration graph, the frame image of selection is as shown in Figure 3.It is tied using processing is obtained after sparse denoising
Fruit figure, as shown in Figure 4.
By a series of noise-containing tennis service figure construction dictionaries acquired, reconstructed according to rarefaction representation glitch-free
The concrete operations of service figure are as follows:
2-1 initialization: generate X=Y, initial dictionary D by DCT；
Above-mentioned X indicates the image after removal noise to be solved, its initial value is set as Y, Y is figure polluted by noise
Picture, the size of X, Y be N × N × 3 (3 indicate that the three primary colors of color images are red, green, the corresponding dimension of blue submatrix,
X, Y is color image), Y is divided into M image subblockSubimage block zjSize isAnd image subblock meets the sparse model z of triple (ε, s, D)j=D α, i.e., it is sub
Image block zjReconstruction error is ε, and code coefficient degree of rarefication is s, and dictionary is D ∈ R3n×K, x ∈ R3nFor signal z to be processedjAgain it arranges
The dimensional vector arranged, α are zjRarefaction representation on dictionary D.
Subimage block zjDenoising image block estimation model for example shown in formula (1):
μ in formulajFor the weight for controlling degree of rarefication and reconstruction error.
Formula (1) is modified slightly to obtain shown in the denoising Image estimation model such as formula (2) of entire image:
X is the clean image (image after denoising) recovered, sub-block x in formula (2)ij=RijX, Rij∈Rn×NBe
The matrix extracted in (i, j) a image block.Similarity between the clean figure of its recovery of first item constraint and noise pattern, Section 2
Prior-constrained for degree of rarefication, Section 3 is the constraint of sub-block reconstruction error.
2-2 fixes dictionary D, and changing formula (2) is formula (3), solves the code coefficient of each sub-block:
The factor alpha that 2-3 fixing step 2-2 is solvedij, formula (2) is turned to formula (4), dictionary D is solved:
Dictionary D is solved using K-SVD algorithm: first finding atom d in Dl(l=1 .., k) coefficient of correspondence αijUnder nonzero element
Mark set ωl=(i, j) | αij(l) ≠ 0 indexed set ω }, is then calculatedlIn each descend target residual error, such as formula (5):
By residual error setColumn vector constitute residual matrix El, SVD is carried out to it is decomposed into U Δ VT.That is U
One column update atom dl, V first row is multiplied with first singular value of Δ updates coefficientFinally obtain adaptive word
Allusion quotation D.
2-4 restores clean image X according to formula (6):
Step 3: the service figure obtained for step 2 carries out sport foreground extraction using mixed Gaussian background modeling, extracts
Motion foreground picture afterwards is as shown in Figure 5.
A, b and c of Fig. 5 is respectively the processing of the 5th frame, the 201st frame, the 575th frame of selected group tennis service video image
Effect.
Step 4: going out the mark point of binaryzation by color feature extracted, and dot profile lookup is marked, use smallest circle
Profile is surrounded, returns to central coordinate of circle as body joint point coordinate, concrete operations are as follows:
The rgb value of each of the foreground image that 4-1 traversal step 3 obtains pixel, for meeting mark point color
Being retained in range, obtains binary image f；
4-1-1 calculates the three primary colors pixel value of all pixels point in foreground image: I=2R-G-B, and will calculate
The value I of acquisition saves as new matrix a Z, i.e., new image Z；
The average value of all I is denoted as A in 4-1-2 calculating matrix Z；
All pixel values greater than A are set to 255 in 4-1-3 matrix I, and all pixel values less than or equal to A are set to 0, obtain
Binary image f；
4-2 carries out profile to bianry image f and searches acquisition contour images f1；
4-3 is to image f1The radius for obtaining the profile point for separately including A, B, C tri- labels by Hough transform is the smallest
Three circles, the center of circle of these three circles are required mark point.
The binary map and profile diagram of acquisition are as shown in a and b of Fig. 6.
Step 5: it is trained using body joint point coordinate of the support vector machines to extraction, realizes the classification to service track,
Assessing tennis service is successfully to serve a ball or unsuccessfully serve a ball, specific as follows:
5-1 distinguishes the X, Y coordinates of mark point of video A, B, C tri- labels of sequential storage service in each frame image, deposits
The coordinate of storage is the running track of mark point；
5-2 obtains N+M (N >=150, M >=50) by step 1 to step 4 and organizes motion profile, wherein service successful trail Ni
+MiGroup, service failure motion profile Nj+MjGroup, Ni+Nj=N, Ni> Nj, Mi+Mj=M, Mi> Mj；
5-3 is trained support vector machines using the N group data of step 5-2, then with the M group data of step 5-2 into
Row test, if service pass flag is 1, service fail flag is -1, completes the classification of service track.
Claims (4)
1. a kind of ball service behavior rating method based on image procossing, which comprises the following steps:
Step 1: color mark being carried out to the artis of service arm, acquires ball service image；
In step 1, color mark is carried out to the artis of service arm, acquiring ball service image, specific step is as follows:
1-1 selects the red marker color as artis, and in three points A, B and the C of label on arm that swing the bat, it is close that A represents wrist
The mark point of racket part, B represent the mark point in elbow portion, and C represents service arm close to the mark point of shoulder；
High-speed camera is placed on the right side of experimental subjects by 1-2, is highly acquisition ball at 1.0m~1.6m apart from penalty mark 4m~6m
Class service image；
Step 2: the ball service image configuration dictionary acquired by step 1 reconstructs glitch-free service figure；
Step 3: the service figure obtained for step 2 carries out sport foreground extraction；
Step 4: extracting mark point for the sport foreground after step 3 operation, and dot profile lookup is marked, use smallest circle
Profile is surrounded, returns to central coordinate of circle as body joint point coordinate；
In step 4, mark point is extracted for the service figure after step 3 operation, and dot profile lookup is marked, use smallest circle
Profile is surrounded, returning to central coordinate of circle, specific step is as follows as body joint point coordinate:
The rgb value of each of the foreground image that 4-1 traversal step 3 obtains pixel, for meeting mark point color gamut
Interior is retained, and binary image f is obtained；
In step 4-1, the rgb value of each of foreground image that traversal step 3 obtains pixel, for meeting mark point face
Being retained in color range, obtaining binary image f, specific step is as follows:
4-1-1 calculates the three primary colors pixel value of all pixels point in foreground image: I=2R-G-B, and obtains calculating
Value I save as new matrix a Z, i.e., new image Z；
The average value of all I is denoted as A in 4-1-2 calculating matrix Z；
All pixel values greater than A are set to 255 in 4-1-3 matrix I, and all pixel values less than or equal to A are set to 0, obtain two-value
Change image f；
4-2 carries out profile to bianry image f and searches acquisition contour images f1；
4-3 is to image f1The profile point for separately including A, B, C tri- provided in step 1-1 labels is obtained by Hough transform
The smallest three circles of radius, the center of circle of these three circles are required mark point；
Step 5: being trained by the body joint point coordinate data that step 1~4 are extracted, realize the classification to service track, evaluation
Ball service is successfully to serve a ball or unsuccessfully serve a ball out, establishes assessment system；
Step 6: service coordinate to be measured being acquired by the method for step 1~4, the assessment system that input step 5 obtains is commented
It is fixed.
2. the ball service behavior rating method based on image procossing as described in claim 1, which is characterized in that in step 2,
A series of noise-containing ball service figure construction dictionaries acquired by step 1, reconstruct glitch-free hair according to rarefaction representation
Ball figure.
3. the ball service behavior rating method based on image procossing as described in claim 1, which is characterized in that in step 3,
Sport foreground extraction is carried out using mixed Gaussian background modeling.
4. the ball service behavior rating method based on image procossing as described in claim 1, which is characterized in that in step 5,
The body joint point coordinate extracted to step 4 is trained, and realizes the classification to service track, and assessing ball service is successfully to serve a ball
Or the specific steps of failure service are as follows:
5-1 distinguishes the X, Y coordinates of mark point of video A, B, C tri- labels of sequential storage service in each frame image, storage
Coordinate is the running track of mark point；
5-2 obtains N+M group motion profile by step 1 to step 4, wherein service successful trail Ni+MiGroup, service unsuccessfully move
Track Nj+MjGroup, Ni+Nj=N, Ni> Nj, Mi+Mj=M, Mi> Mj, N >=150, M >=50；
5-3 is trained support vector machines using the N group data of step 5-2, is then surveyed with the M group data of step 5-2
Examination, if service pass flag is 1, service fail flag is -1, completes the classification of service track.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710100439.5A CN106964137B (en) | 2017-02-23 | 2017-02-23 | A kind of ball service behavior rating method based on image procossing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710100439.5A CN106964137B (en) | 2017-02-23 | 2017-02-23 | A kind of ball service behavior rating method based on image procossing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106964137A CN106964137A (en) | 2017-07-21 |
| CN106964137B true CN106964137B (en) | 2019-05-28 |
Family
ID=59328286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710100439.5A Active CN106964137B (en) | 2017-02-23 | 2017-02-23 | A kind of ball service behavior rating method based on image procossing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106964137B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107812343B (en) * | 2017-08-29 | 2019-12-03 | 浙江理工大学 | A kind of vault sports training method |
| CN107670258B (en) * | 2017-11-03 | 2020-09-04 | 杭州乾博科技有限公司 | Table tennis match-up service identification system and method based on positioning |
| CN109045662B (en) * | 2018-08-27 | 2021-03-02 | 李泽栋 | Vision-based badminton serving referee auxiliary system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101982836A (en) * | 2010-10-14 | 2011-03-02 | 西北工业大学 | Mark point identification initializing method based on principal components analysis (PCA) in motion capture system |
| CN103295016A (en) * | 2013-06-26 | 2013-09-11 | 天津理工大学 | Behavior recognition method based on depth and RGB information and multi-scale and multidirectional rank and level characteristics |
| CN105536205A (en) * | 2015-12-08 | 2016-05-04 | 天津大学 | Upper limb training system based on monocular video human body action sensing |
| CN105844258A (en) * | 2016-04-13 | 2016-08-10 | 中国农业大学 | Action identifying method and apparatus |
-
2017
- 2017-02-23 CN CN201710100439.5A patent/CN106964137B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101982836A (en) * | 2010-10-14 | 2011-03-02 | 西北工业大学 | Mark point identification initializing method based on principal components analysis (PCA) in motion capture system |
| CN103295016A (en) * | 2013-06-26 | 2013-09-11 | 天津理工大学 | Behavior recognition method based on depth and RGB information and multi-scale and multidirectional rank and level characteristics |
| CN105536205A (en) * | 2015-12-08 | 2016-05-04 | 天津大学 | Upper limb training system based on monocular video human body action sensing |
| CN105844258A (en) * | 2016-04-13 | 2016-08-10 | 中国农业大学 | Action identifying method and apparatus |
Non-Patent Citations (2)
| Title |
|---|
| 基于混合高斯改进算法的视频背景建模研究;赵子博;《中国优秀硕士学位论文全文数据库信息科技辑》;20140531;全文 * |
| 基于视频图像分析技术的网球发球技术训练的研究;孟号翔;《文体用品与科技》;20141130;第123-124页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106964137A (en) | 2017-07-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | [Retracted] Intelligent Sports Training System Based on Artificial Intelligence and Big Data | |
| CN107423730B (en) | Human gait behavior active detection and recognition system and method based on semantic folding | |
| CN113762133B (en) | Dead weight body-building auxiliary training system, method and terminal based on human body gesture recognition | |
| CN106022213B (en) | A kind of human motion recognition method based on three-dimensional bone information | |
| WO2017133009A1 (en) | Method for positioning human joint using depth image of convolutional neural network | |
| CN108052896A (en) | Human bodys' response method based on convolutional neural networks and support vector machines | |
| CN102074034B (en) | Multi-model human motion tracking method | |
| CN110991340B (en) | Human body action analysis method based on image compression | |
| CN108334899A (en) | Quantify the bone age assessment method of information integration based on hand bone X-ray bone and joint | |
| CN110222665A (en) | Human motion recognition method in a kind of monitoring based on deep learning and Attitude estimation | |
| CN110738154A (en) | pedestrian falling detection method based on human body posture estimation | |
| CN106964137B (en) | A kind of ball service behavior rating method based on image procossing | |
| CN105536205A (en) | Upper limb training system based on monocular video human body action sensing | |
| CN110490109B (en) | Monocular vision-based online human body rehabilitation action recognition method | |
| CN106815855A (en) | Based on the human body motion tracking method that production and discriminate combine | |
| CN112668531A (en) | Motion posture correction method based on motion recognition | |
| CN111833439B (en) | Artificial intelligence based ammunition throwing analysis and mobile simulation training method | |
| CN114618142A (en) | Auxiliary training system and method for table tennis sports | |
| CN115661856A (en) | User-defined rehabilitation training monitoring and evaluating method based on Lite-HRNet | |
| CN113283373A (en) | Method for enhancing detection of limb motion parameters by depth camera | |
| CN103310191A (en) | Human body action identification method for motion information imaging | |
| Liu et al. | Key algorithm for human motion recognition in virtual reality video sequences based on hidden markov model | |
| CN116993760A (en) | Gesture segmentation method, system, device and medium based on graph convolution and attention mechanism | |
| CN114639168B (en) | Method and system for recognizing running gesture | |
| CN113673327B (en) | Penalty hit prediction method based on human body posture estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201016 Address after: 226200 No. 46 Industrial Park, Nanyang Town, Nantong, Jiangsu, Qidong Patentee after: Qidong plant medical equipment factory Address before: 310018, No. 928, No. 2, Poplar Street, Hangzhou economic and Technological Development Zone, Zhejiang Patentee before: Zhejiang University of Technology |